RFR: 8330969: scalability issue with loaded JVMTI agent [v2]

[Daniel D. Daugherty]

On Wed, 15 May 2024 06:00:46 GMT, Serguei Spitsyn <sspitsyn at openjdk.org> wrote:

>> I'm not sure this answered Chris' query properly. Or I'm reading Chris' query wrong.
>> 
>> Perhaps this is not what Chris had in mind, but I'm wondering what happens in some
>> Thread-A when it is checked and passed by but then Thread-A sets the flag in itself
>> after the for-loop has passed it by. Does that Thread-A flag value get lost?
>
>> Perhaps this is not what Chris had in mind, but I'm wondering what happens in some
>> Thread-A when it is checked and passed by but then Thread-A sets the flag in itself
>> after the for-loop has passed it by. Does that Thread-A flag value get lost?
> 
> Thank you for the question.
> The Thread-A sets the flag optimistically and then re-checks if `sync_protocol_enabled()` and any disabler exists. It can be global disbaler (`_VTMS_transition_disable_for_all_count > 0`) or disabler of `Thread-A` only (`java_lang_Thread::VTMS_transition_disable_count(vth()) > 0`). If any disabler exists then `Thread-A` clears the optimistic settings and goes with the pessimistic approach under protection of `JvmtiVTMSTransition_lock`.
> 
> Please, let me know if you still have questions.

This algorithm sounds correct. Thanks for closing the loop on my belated comment.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/18937#discussion_r1603957324